JP2675529B2 - Video coding method and apparatus thereof - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a video signal coding method and apparatus, and for example, in order to code a video having a different complexity at a constant bit rate, a compression integer depending on the complexity is obtained to quantize the video. In this case, the present invention relates to an optimal video encoding method and its apparatus.

[0002]

2. Description of the Related Art Storage and communication technology is rapidly developing due to the recent development of computer, semiconductor and digital signal processing technology. Along with this, digital video devices, such as digital VTRs, digital video cameras, videophones or television phones, are required to have a technique for efficiently recording video information on a recording medium and improving the quality of reproduced video.

However, when a video signal is recorded or transmitted in a digital system, the data amount of the digital video signal is much larger than that in the analog system.
In order to use the recording medium and the transmission channel efficiently, it is necessary to compress the video information. The methods mainly used in the past in connection with video signal compression are the discrete cosine transform technique and the variable length coding technique. FIG. 5 is a block diagram showing a configuration of a conventional video signal encoding device. The discrete cosine transform unit 11 has 64 video blocks of a certain size, for example, 8 × 8 video blocks, which are time domain data. Convert to discrete cosine transform coefficient in the frequency domain of.

The quantizing unit 12 quantizes the discrete cosine transform coefficient output from the discrete cosine transform unit 11 at a predetermined quantizing interval. The variable-length coding unit 13 allocates a small number of bits to data having a small amount of information, that is, data having a low information value, and allocates a large number of bits to data having a large amount of information, that is, data having a high information value. Coded data of variable length is generated so as to be allocated, and the number of bits required for coding is minimized. Since the length of the data output from the variable length encoding unit 13 is not constant, the buffer 14 temporarily stores the data and then outputs the data at a constant bit rate.

Conventionally, the discrete cosine transform unit 11 divides the input image into small sub-images of 8 × 8 blocks and processes them independently, and the quantizing unit 12 processes the discrete cosine transform coefficients in the buffer 14 and the current state. Quantize at the quantization interval determined by the complexity of the input image. Therefore, if the difference between the quantization intervals between adjacent macroblocks is large, the correlation between the blocks is lost at the boundary surface of the adjacent macroblocks, and a so-called block effect appears.

In addition, when quantizing each macroblock at a fixed quantization interval, since the complexity of each macroblock is different even within one screen, errors occur unevenly on the screen and a specific portion is generated. Artifacts that reduce the image quality of the image appear, and a suitable quantization interval should be determined so that such an artifact does not occur, and this determination is important above all.

On the other hand, in the variable-length coding, the length of the generated code is variable, so that the number of generated bits must be controlled to be constant in order to reduce the reproduction during special reproduction such as high-speed search. The compression integer, which is an important parameter that determines the bit amount, determines the quantization interval of the quantizer, which determines the compression ratio and also affects the screen resolution.

[0008]

However, as shown in FIG. 5, when a buffer is used to control the bit amount to be constant, the quantization interval is controlled by the fullness of the buffer, so that it is possible to change the quantization interval within one screen. There is a problem that it is difficult to allocate bits suitable for a specific area (macroblock) according to complexity.

On the other hand, after detecting the complexity of the DCT transform block, a video signal coding / decoding system which alleviates the blocking phenomenon by adaptively adjusting the quantizing interval of the quantizer according to the relationship of the surrounding DCT transform blocks. Are disclosed in US Pat. No. 5,253,075. Although the above-described device detects the complexity of an image and adjusts the quantization interval by using the information in the frequency domain, the conventional video signal encoding / decoding system uses a predetermined block (macro) in the spatial domain. It was not possible to provide a means for encoding a block) at a constant bit rate.

An object of the present invention is to provide a video coding which determines a compression integer capable of coding video data at a constant bit rate by adapting to the complexity detected in a sub-block unit composed of a predetermined number of unit blocks. Is to provide a method.
Another object of the present invention is to provide a video encoding method that adjusts the quantization interval in accordance with the complexity detected in the spatial domain and the frequency domain in units of macroblocks.

Still another object of the present invention is to provide a video coding apparatus for determining a compression integer by adapting to the complexity detected in sub-block units and determining a quantization interval.

[0012]

In order to achieve the above object and other objects, a video encoding method according to the present invention has the following configuration, for example. That is, a discrete cosine transform unit that transforms a sub block, which is data in the time domain, into a discrete cosine transform coefficient, which is data in the frequency domain, with respect to an input image divided into sub block units configured by a predetermined number of unit blocks. , A video coding comprising a quantizer for quantizing the discrete cosine transform coefficient at a predetermined quantization interval determined by a compressed integer, and a variable length encoder for variable length coding the quantized coefficient. In the apparatus, a complexity calculation unit that calculates complexity in pixel units of an input image in units of the sub-blocks, a complexity normalization unit that normalizes the complexity for each sub-block, and the discrete cosine transform unit. A zero coefficient counting unit that counts small zero coefficients within a predetermined range in units of the sub-blocks by the discrete cosine transform coefficients converted by the above, and normalizes the zero coefficients for each sub-block. A zero coefficient normalization unit, a weight value calculation unit that calculates a transition weight value with the zero coefficient and a preset compression ratio, and the complexity normalization value, the zero coefficient normalization value and the transition weight value are added. And a compressed integer generator for generating the compressed integer and outputting the compressed integer to the quantizer.

Further, for example, the sub-block is a macro block unit. Alternatively, the predetermined number in the zero coefficient counting unit is ± 1.

[0014]

In the above structure, the compression interval to be coded at a constant bit rate is calculated according to the complexity of the input image in the spatial domain and the complexity in the frequency domain to adjust the quantization interval of the quantizer. In addition, it is possible to perform coding without deterioration of image quality even in a screen composed of images of different complexity.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail below with reference to the accompanying drawings. 1A to 1C are
An image in the 4: 2: 2 format of the CCIR601 recommendation is shown, and A of FIG. 2 and B of FIG. 2 are drawings for explaining the configuration of the macro block for the Y (circuit) screen.

Each image is composed of Y (A in FIG. 1), Cr (B in FIG. 1) and Cb (C in FIG. 1) signals, and each signal is divided into macroblocks. Each macroblock contains four 8x8 pixel block signals, Cr and Cb, for the Y signal.
Each signal consists of two 8 × 8 pixel block signals. A of FIG. 2 shows one screen of the Y signal, and B of FIG. 2 shows a macro block of 16 × 16 pixels for the Y signal.

FIG. 3 is a flowchart of a video encoding method for determining a compressed integer according to an embodiment of the present invention.
The video encoding method of this embodiment for determining a compression integer for an input image will be described in detail with reference to FIGS. 1A to 2B. The definitions of the symbols used in the mathematical formulas are as follows. Pix (k, l) is a pixel value, and k is 0, 1, ...
, 7, l is 0, 1, ..., 7, N is the number of blocks (N = 4 in this embodiment), MB is the number of macroblocks (MB = 30 × 45 in this embodiment), and i is 0 , 1, 2, 3, ...,
MB-1, j is 0, 1, ..., N-1.

As a first step, in order to obtain the complexity normalization value for each macroblock according to the complexity for the spatial domain, the input 4: 2: 2 format image (step 410) is performed in the macroblock. Each 8x8
The average complexity of the image is calculated in block (hereinafter referred to as unit block) units (step 421). The complexity is calculated for the luminance signal Y by using the pixel value of the image.

The average complexity A _yb [j] of the unit block is obtained by the following equation (1).

[0020]

(Equation 1)

After summing up all the average complexity of the unit blocks by the following equation 2 for the macroblock,
Average complexity A of macroblocks divided by the number of unit blocks
_{Find yt} [i] (step 422).

[0022]

(Equation 2)

Here, N is 4. Then, the root mean square value σ ² _BT [j] of the unit block with respect to the difference between the average complexity of each unit block of the macroblock and the actual pixel value is calculated by the equation (3) (step 423).

[0024]

(Equation 3)

Then, according to the equation 4, 4 of each macroblock
For one Y unit block (j = 0 to 3), the least mean square value of the mean square values of each unit block is calculated, and then 1 is added, and the least mean square value σ ² _{BT min of} each macro block is added.
Find [i] (step 424).

[0026]

## _EQU4 ## σ ² _BTmin [i] = MIN (σ ² _BT [j]) +1 Here, 1 is added so that the least mean square value does not become "0". Then, each macroblock is normalized by the least square mean value of the macroblock and the average complexity according to the equation (5) to obtain a macroblock complexity normalization value Nor _ACT [i] (step 425).

[0027]

(Equation 5)

Here, α is a constant for normalization.
As a second step, in order to obtain the zero coefficient normalization value and the transition weight value for each macroblock in the frequency domain of the video signal, the input 4 based on the equations (6) and (7):
The coefficient (Pix _{Coef f (u, v)} ) obtained by performing the discrete cosine transform on the 2: 2 format image 41 is obtained (step 431).

[0029]

(Equation 6)

Here, u, v = 0, 1, ..., 7 are satisfied.

[0031]

(Equation 7)

Next, according to the equation (8), Y of the macroblock is
Count all coefficients with discrete cosine transform coefficients smaller than ± 1 for a block and sum zero coefficients of macroblocks Z
Find [i] (step 432).

[0033]

(Equation 8)

Then, the zero coefficient sum of each macroblock is added to all the macroblocks by the formula (9) to obtain the total zero coefficient sum Z _T (step 433).

[0035]

(Equation 9)

Then, the total zero coefficient sum is divided by the number of macroblocks by the expression (10) to obtain the average zero coefficient sum A _ZERO of the macroblocks (step 434).

[0037]

## EQU10 ## A _ZERO = (Zr / MB) Subsequently, the difference between the average zero coefficient sum of the macroblocks and the zero coefficient sum of each macroblock is squared by the equation 11 to calculate the zero coefficient square mean value σ ^{2 of the} macroblocks. _{Calculate ZERO} [i] (step 435).

[0038]

[Equation 11] σ ² _ZERO [i] = (A _ZERO −Z [i]) ² Next, the zero coefficient per macroblock is normalized by the formula 12 and the zero coefficient of each macroblock is normalized. Obtain the normalized value Nor _ZERO [i] (4th
36 steps).

[0039]

(Equation 12)

Here, α is a constant for normalization.
The compression integer varies depending on the complexity of the image, and the quantization interval is determined by the compression integer. The quantization interval can be changed by the frequency of occurrence of the zero coefficient, and the compression integer is controlled by the compression ratio. Therefore, if W is a transition weight value of the compression integer, it can be expressed as follows.

[0041]

Equation 13] W = (A _ZERO chromatography (β × RATIO)) / γ Accordingly, finding a transient weight value W for each macroblock in accordance with the compression ratio to be compressed between the average zero coefficient sum of macro blocks (No. 441 Stage). Where β, γ
Is a constant for weighting value, RATIO is a compression rate, and the compression rate is predetermined.

Therefore, the compression integer SF [i] in macroblock units is determined by the sum of the complexity normalization value, the zero coefficient normalization value, and the transition weight value according to the following Equation 14 (step 450).

[0043]

SF [i] = Nor _ACT [i] + Nor _ZERO [i] + W The compressed integer determines the quantization interval of the discrete cosine transform coefficient by the quantization matrix in the quantizer. That is, the larger the sum of the average zero coefficients of macroblocks and the smaller the compression rate, the larger the transition weight value W. The larger the transition weight value W, the larger the compression integer and the smaller the quantization interval.

FIG. 4 is a block diagram showing an example of the arrangement of a video coding apparatus capable of executing the above-mentioned video coding processing according to an embodiment of the present invention. The encoding device of the present embodiment uses Y, C
For an input image divided into macroblocks of r and Cb signals, a discrete cosine transform unit 51 and a discrete cosine transform unit 51 that transform macroblock data that is time domain data into discrete cosine transform coefficients that are data in the frequency domain. Quantizer 58 that quantizes the discrete cosine transform coefficient output from the unit at a predetermined quantization interval, variable that generates coded data of variable length according to the amount of information and coded with a minimum number of bits as a whole. The conventional video coding apparatus having the long coding unit 59 further includes a configuration for obtaining a compressed integer for the input video.

Next, FIG. 4 will be described focusing on the configuration added in comparison with the conventional device shown in FIG.
The complexity calculator 52 calculates the complexity according to the pixel value of the input image, and the complexity normalizer 53 normalizes the complexity for each macroblock. Zero coefficient counting unit 54
Counts zero coefficients smaller than ± 1 among the discrete cosine transform coefficients converted by the discrete cosine transform section 51, and the zero coefficient normalization section 55 normalizes the zero coefficient for each macroblock.

The weight value calculation unit 56 calculates the transition weight value for the macroblock based on the zero coefficient counted by the zero coefficient counting unit 54 and the compression rate. The adder 57 adds the complexity normalization value, the zero coefficient normalization value, and the transition weight value to generate a compressed integer and outputs the compressed integer to the quantizer 58. The quantization interval for the discrete cosine transform coefficient is determined by the compression integer and the quantization matrix. The degree of change of the quantization matrix is different for every input image, the characteristics of the color components, the resolution, the application purpose, etc., so it is necessary to find a quantization matrix suitable for it.
It is obtained by various psychovisual experiments that combine the human visual characteristics and the characteristics of the discrete cosine transform coefficient.

The quantizer 58 quantizes the video signal according to the determined quantization interval, and the resulting value is the variable length coder 5.
A compressed bitstream of desired bit rate is obtained via 9. Although the macroblocks are coded at a constant bit rate in the above-described embodiment, the macroblocks may be expanded in block units.

[0048]

As described above, according to the present invention, a compressed integer which is encoded at a constant bit rate depending on the complexity of the input image in the spatial domain and the complexity in the frequency domain (frequency of occurrence of zero coefficient). By adjusting the quantization interval of the quantizer in order to obtain, it is possible to perform coding without deterioration of image quality even in a screen composed of images of different complexity.

[Brief description of the drawings]

FIG. 1 is a CCIR 601 4: 2: 2 format diagram of an input image.

FIG. 2 is a diagram showing a configuration map for the Y signal shown in FIG.

FIG. 3 is a flowchart showing a process of determining a compressed integer by the video encoding method according to the embodiment of the present invention.

FIG. 4 is a block diagram showing a configuration of a video encoding device according to an embodiment of the present invention.

FIG. 5 is a block diagram showing a configuration of a conventional video encoding device.

[Description of Signs] 52 Complexity Calculation Unit 53 Complexity Normalization Unit 54 Zero Coefficient Counting Unit 55 Zero Coefficient Normalization Unit 56 Weight Value Calculation Unit 57 Adder 58 Quantizer 59 Encoder

Claims (11)

(57) [Claims] 1. An image to be coded after dividing an input image into sub-block units composed of a predetermined number of unit blocks and quantizing a divided video signal into a predetermined quantization interval determined by a compression integer. In the encoding method, a step of obtaining a normalized value of the complexity for each sub-block according to the complexity of the input image in the spatial domain, and a discrete cosine transform for the input image, and for each sub-block A step of counting zero coefficients whose discrete cosine transform coefficients are smaller than a predetermined number to obtain a zero coefficient sum; a step of obtaining a zero coefficient normalized value for each sub-block by the zero coefficient sum; and the zero coefficient sum being preset. Determining a transition weight value for each sub-block according to a compression rate, adding the complexity normalization value, the zero coefficient normalization value, and the transition weight value to obtain the transition weight value. Image encoding method which comprises a step of determining a contraction integer. 2. The video encoding method according to claim 1, wherein the complexity normalization value is calculated for the luminance signal in the complexity normalization value calculating step. 3. The video encoding method according to claim 1, wherein the predetermined number is ± 1. 4. The video encoding method according to claim 3, wherein the coefficient for which the discrete cosine transform coefficient is smaller than ± 1 in the step of calculating the zero coefficient value is a coefficient for Y. 5. The video encoding method according to claim 1, wherein the sub-block is a macro block. 6. The step of calculating the normalized complexity value is a step of obtaining an average complexity of each unit block of the macroblock, and an average complexity of the macroblock is obtained according to the average complexity of the unit block. The step of calculating the difference between the average complexity of the unit block and the actual pixel value to obtain the root mean square value of the unit block, and the macro block to the minimum value of the root mean square value for each unit block. 1 is added to obtain the least square mean value of the macroblocks, and each macroblock is normalized according to the least square mean value of the macroblocks and the average complexity to obtain the complexity normalization value of each macroblock. The video encoding method according to claim 5, further comprising the steps of: 7. The step of calculating the zero coefficient normalization value includes a step of summing all zero coefficient sums per macroblock to obtain a total zero coefficient sum, and an average zero of macroblocks according to the total zero coefficient sum. A step of obtaining a coefficient sum, a step of obtaining a zero coefficient root mean square value of the macroblock according to a difference between the average zero coefficient sum of the macroblock and the zero coefficient sum of the macroblock, and a zero coefficient root mean square value of the macroblock And normalizing according to the sum of zero coefficients to obtain a zero coefficient normalized value of each macroblock.
The video encoding method described in the item. 8. The transition weight value calculating step includes a step of summing all zero coefficient sums of the macroblocks to obtain a total zero coefficient sum, and a step of obtaining an average zero coefficient sum of the macroblocks by the total zero coefficient sum. 6. The video encoding method according to claim 5, further comprising: determining a transition weighting value for each macroblock based on a sum of average zero coefficients of the macroblocks and a predetermined compression rate. 9. A discrete cosine for converting a sub-block, which is data in the time domain, into a discrete cosine transform coefficient, which is data in the frequency domain, with respect to an input image divided into sub-block units composed of a predetermined number of unit blocks. A transforming unit, a quantizing unit for quantizing the discrete cosine transform coefficient at a predetermined quantizing interval determined by a compressed integer, and a variable length coding unit for variable length coding the quantized coefficient. In the video encoding device, a complexity calculation unit that calculates a complexity by a pixel value of an input image in units of the sub-blocks, a complexity normalization unit that normalizes the complexity for each sub-block, and the discrete A zero coefficient counting unit that counts small zero coefficients within a predetermined range in the sub-block unit by the discrete cosine transform coefficient converted by the cosine conversion unit, and the zero coefficient for each sub-block. A zero coefficient normalizing unit for normalizing, a weight value calculating unit for calculating a transition weight value by the zero coefficient and a preset compression ratio, the complexity normalization value, the zero coefficient normalizing value and the transition weight And a compressed integer generator that adds a value to generate the compressed integer and outputs the compressed integer to the quantizer. 10. The video encoding apparatus according to claim 9, wherein the sub-block is a macro block unit. 11. The predetermined number in the zero coefficient counting unit is ±
The video encoding device according to claim 9, wherein the video encoding device is 1.